selectparks - Nero: Predatory Play

Students and reseachers at the University of Texas, Austin, have built a system that uses gameplay as a means of training software robots that can then be deployed in game. Using 'machine learning' techniques that interface with a downloadable client game, bots are schooled to deal with a variety of different situations ensuring they are actively responsive agents when engaged in the next turn of combat.

Aside from obviously unfortunate applications in training corporeal war machines (ask the boss, it's Texas afterall), 'Nero' does raise questions as to what other areas this technique could be applicable. Perhaps we'll be play-training our houses, blimps, computers and vehicles how to respond to a variety of hypothetical circumstances in the future.

Maybe they'll call this 'Preparatory Play' or something.

Resources:
Game AI Links